October 1996 € Volume 23 € Number 5

Practice Lessons
OR Boosts AT&T Capital's Leasing Business

By David Greenfield

The Edelman competition, sponsored by the College on the Practice of Management Science (CPMS), the Practice Section of INFORMS, annually recognizes and rewards outstanding achievement in the practice of operations research and the management sciences. Persons interested in more information on this or other recent Edelman work are encouraged to read the annual special issue (January/February) of Interfaces in which the full papers submitted in conjunction with the Edelman competition are published. To obtain a copy of the most recent special issue of Interfaces or for subscription information to the journal, call 1-800-4INFORMs.

AT&T has been recognized for years as a standard-bearer in the use and promotion of operations research. In 1994, the company received the ORSA Prize for its commitment to operations research. Now, at AT&T Capital Corporation's small-ticket leasing business, operations research has been put to use in the design of three sets of decision automation systems and associated strategies for front-end credit decisioning, life cycle credit line management, and delinquent account collections.

The small-ticket leasing business involves all varieties of equipment leasing where the transaction purchase is less than $50,000. According to Gary Kochman, technical manager of the Operations Research Models Group at AT&T Labs, "The types of equipment AT&T Capital leases are organized into different strategic business units, focusing on different equipment markets. AT&T started this business in 1985 to provide attractive leasing and financing terms to support the sales of AT&T telecommunications equipment."

To underscore the impact that leasing has had on American business today, the authors of the AT&T credit and collections decision automation project quote U.S. Department of Commerce figures stating that 80 percent of companies today lease all or some of their equipment. The reason so many companies do this is because the value of equipment comes from its use, not its ownership. Of the estimated $571 billion spent by business on productive assets in 1995, $161 billion (28 percent) was acquired through leasing. American leasing companies own over $400 billion in equipment, more than any other industry [U.S. Department of Commerce 1995].


Starting the project
AT&T Capital, along with AT&T Bell Laboratories' Operations Research Department, began the development of the three decision automation systems for the small-ticket leasing business in 1992, following the installation of the first credit scoring system to automate the credit decisioning process at AT&T Leasing Services. "They (AT&T Leasing) asked us to come in and review and validate their system and look for improvement opportunities. Out of that analysis came recommendations on some immediate steps they could take to improve the quality of the decisions they were making as a result of the system, and some longer-term recommendations on steps they could take to improve the quality of the risk models that were driving the system," Kochman says.

"AT&T was looking for decision automation to increase the productivity of their credit granting decisions and to reduce costs. The company also wanted to improve the quality of its portfolio. In addition, AT&T was looking for quicker response times to its customers."

In a vendor-based leasing business, the customer is present at the vendor location at the time of the loan application. So when the vendor faxes that customer's leasing application to two or three leasing companies, the first company to respond positively usually captures the business. "Automation greatly reduces this response time and helped AT&T to increase its market share in this area. In retrospect, that's probably the largest benefit we got out of the systems by far," Kochman says.

The front-end credit systems portion of the project utilizes a two-component methodology &emdash; one of which is relatively conventional and the other is new, claims Kochman. The first component is risk prediction, where models are built based on historical data relating to customers' past actions detailed in AT&T's books. "We developed models from this to predict the future performance of new applicants in order to try and identify which ones were credit worthy and which ones were too great a risk for us to warrant extending the kind of credit the leasing transaction would involve. We used a very conventional risk prediction methodology. I think we were innovative in the performance metric that we used, but the underlying methodology itself was conventional risk prediction," he says. "Once we could predict risk, however, we still had a question regarding decision making, and that had two levels in itself &emdash; 1) Where do we go to get information that could be used to drive the risk prediction? and 2) Once we have information from a few sources, how do we decide which source to use?"

Several organizations exist for the capture of such information, including credit bureaus and consumer bureaus. "Where you go to get the data can impact the cost of acquiring the data and the quality of the prediction itself," Kochman claims. "We basically built a decision logic to drive decisions about which source of data to access first, and to drive decisions about whether that data was enough to make a decision or whether we needed to go to a second source for supplemental information. Then, based on the information we collected, the second level of decision logic involved the approval, rejection or manual review of the leasing request. Based on the information we were able to get, the system either automatically approved or rejected the application, or decided it couldn't decide by itself and sent it to a credit manager for manual review. We then moved into the design of a decision logic model to decide which bureau to use under what circumstances. The decision flow logic model used for this portion of the system is new to the industry. The benefit we receive from this model is that we know we are making optimal use of the most economical sources of data."

The credit line management system developed by AT&T is entirely new to the industry. To create this system, the AT&T OR team had to approach the problem differently than the consumer credit card industry where established methodologies already exist. The notion of automating the setting and updating of credit line histories was new to the leasing industry and, therefore, the methodology was also new. The methodology was, however, similar to what was used in the front end (risk prediction) models.

"We had to take the various dynamics of the leasing business into account when developing this model," Kochman says. "When a new customer comes to you for a first deal, all you have available to you is credit bureau information. When you put a transaction on the books, each month you get more payment experience with that customer, and it becomes a question of how long do you rely on the original credit information you received from the credit bureau and when can you start to rely on the information you can glean from the payment experience you've gained from that customer. So a series of models were built, all based on how much payment experience we had with a particular customer."

The collections management system also used conventional risk prediction models, but the models were customized to work on the accounts that became delinquent. To handle the collections arm of AT&T's leasing business, a four-component decision support system was designed. The four parts of this sytem include:

• a risk assessment module,
• collection activity and treatment scenarios,
• assignment of treatment scenario based on risk, and
• automation and prioritization of accounts to customer service representatives.

Implementation and progress
Once the systems were developed, they were fully integrated into AT&T Capital's current systems. The software needed to interface with the bureaus and implement the decision/risk models within AT&T Capital's system was developed and put into place through a combination of AT&T's internal information systems group and, in some cases, contracts with outside vendors.

"A substantial amount of training was involved in making these systems a part of the business' day-to-day operations," Kochman says. "The credit managers were made to realize how the systems worked and what the significance of the scores were so that they understood operationally what information the system provided to them."

According to the AT&T teams' Edelman paper, the implementation of these systems has led to the vast majority (73 percent) of AT&T Capital's small-ticket business credit-granting decisions becoming automated, including the automatic approval of $685 million in new transactions annually. Productivity gains have enhanced competitiveness by reducing response times, and increased profitability by reducing credit and collections operating costs by over $3.1 million annually. In addition, improvements in decision quality through deployment of these systems have led to business volume gains of $86 million annually, while simultaneously reducing bad debt losses by $1.1 million annually.

Steven Meester, principal technical staff member at AT&T Capital Labs, stresses the early involvement of all parties affected by the new systems. "The most important aspect of this project was to involve, very early on in the process, all the stakeholders and people from the information, credit and operation sides of the business."

"Having everyone involved early increases the level of buy-in that you experience at the back end of the project during implementation. And, more importantly, there's a lot to be learned at the front end during the design phase from the people who have been working in this business all along," Kochman concurs.

"Following this type of development process, you learn early on about the limitations of the infrastructure and the systems, so you don't wind up designing a system that can't be implemented," says Kochman. "That's the biggest reason for having operations and information systems people involved from the start.

"And it goes without saying that having high-level management support from start to finish is very important. We had that all along on this project, so it was almost a non-issue. But we benefited a lot from not having to second guess whether this project was worthwhile doing. And knowing that we would have the resources allotted to finish this program assured us that management was committed to seeing this through."

This project's success keeps AT&T headed toward further developments with OR.

"The success of these systems at AT&T Capital has led to an increase in respect for what such techniques can offer and has served as a springboard for ongoing demand for the technology," Kochman adds. "We're not resting on our laurels, either. As other companies learn about what we've done here and begin to adopt the methodologies, by the time they get them into place, we expect to be at the next level."


David Greenfield is the managing editor of OR/MS Today.